Most recent kernel where this bug did not occur: 2.6.12 Distribution: Mandriva 2006 (Cooker) Hardware Environment: Acer Aspire 1705SMi, Intersil Corporation Prism 2.5 Wavelan chipset (rev 01) - internal wireless 802.11a/b card Software Environment: KDE 3.4.2, X.org X11 6.9 patch level 1.cvs20050915.2mdk Problem Description: S3 suspends, but no longer resumes successfully since 2.6.12 - echo mem > /sys/power/state puts it to sleep quickly, but when resumed video stays dead and there are two cases: --Under X: CPU seems under constant load (fan quickly goes up through the levels until its at max) - a few times, video has come back just long enough to see a window popped up displaying a "System is going down for maintenance NOW!!" message, but for less than a second so I couldn't get any other details. --Under console having killed X (and/or shutting down without -h, which forces this machine to drop into single-user mode in sh): Video usually stays dead and system shuts down after about 20 seconds. In both cases, I can at least reboot the machine with magic-sysrq + b. Steps to reproduce: Acquire an Acer Aspire 1705SMi, replace working 2.6.12 kernel with 2.6.13 or 2.6.14-rc2, echo mem > /sys/power/state, then try to wake up machine. Throw laptop through window. Notes: I'll provide any data I can, and try patches if anyone has any ideas. Initially, I assumed it was a problem with orinoco_pci/hermes, the inbuild Intersil pci wireless card was causing problems, as it would echo hundreds of "BAP0 offset error: ..." messages into /var/log/messages and syslog, even having unloaded the modules - but this didn't happen the last couple of times using the latest rc of the driver. Ois
Created attachment 6146 [details] lsmod for pre-suspend system in X
Created attachment 6147 [details] brief lspci for running system
Created attachment 6148 [details] syslog over suspend, resume and (strange) shutdown events
Created attachment 6149 [details] dmesg... spammy
Could you please try a minial config in the system (Eg. remove some device drivers)? >ACPI: PCI Interrupt Routing Table [\_SB_.PCI0.AGPB._PRT] > uteval-0214: *** Error: Return object type is incorrect [\_SB_.PCI0.LPCB.LNKA._CRS] (Node c1471d28), AE_TYPE > uteval-0221 [10] ut_evaluate_object : Type returned from _CRS was >incorrect: Integer, expected Btypes: 4 >pci_link-0286 [07] pci_link_get_current : Error evaluating _CRS Please try not to override the DSDT.
> tbget-0284: *** Info: Table [DSDT] replaced by host OS We can't debug this with a modified BIOS. Please attach the acpidump output for the OEM's latest BIOS, and if you modify that, please attach a source code patch to it. Also, please attach the working 2.6.12 dmesg for comparison. if it has the _CRS error messages too, then we know that isn't the issue and we need to look elsewhere.
Created attachment 6205 [details] dmesg for 2.6.12 boot Here's the dmesg of the 2.6.12 boot and across a suspend and resume from S3 in X. After much effort, I got nowhere with 2.6.14-rc2 in X's S3 support - and had to hack the nvidia driver's source to get it to even try to suspend. The same thing happened upon trying to wake the machine - the video stayed dark and the fans spun higher and higher as the CPU ran itself into the ground. For the first 10 seconds or so, the system was still somewhat alive - hitting the numlock key would result in the LED lighting up about a second later, but soon even this stopped working. I'd rather not try updating the BIOS on this machine - the revision history on the latest release doesn't seem to mention anything that might affect suspend support. Last time I tried updating the Acer's BIOS, it went awry and since Acer didn't design any method of emergency flashing the BIOS or even removing the flash RAM chip (or NVRAM, or whatever it is). Even though the machine was under warranty, I was forced to pay over 200 euro for them to replace the whole motherboard. I felt really cheated over that - so I'm not willing to risk going through that again if there's any way around it. About the DSDT override, I couldn't get battery state information without it, but that was back in 2.6.9 or so. Will try rebuilding the kernel without the override and extracting the bios tomorrow. Thanks for the help, it's much appreciated.
Created attachment 6206 [details] acpidump output ...Or I could just do that now :)
Created attachment 6207 [details] modified dsdt source The modified DSDT I created so I could read battery status and stuff.
Can you access the system remotely to get the dmesg? Also please kill acpid daemon before suspend, and try again.
Please reopen this bug if: - it is still present in recent 2.6 kernels and - you can provide the requested information.